Análisis de ingredientes en recetas

Análisis de https://www.nature.com/articles/srep00196.pdf

Podemos usar read_lines_chunked si el archivo original es grande:

library(tidyverse)
limpiar <- function(lineas,...){
  str_split(lineas, ',') %>% 
    keep(~.x[1] == 'EastAsian') %>%
    map(~.x[-1]) %>% # quitar tipo de cocina
    map(~.x[nchar(.x) > 0]) # quitar elementos vac{ios}
}
filtrado <- read_lines_chunked('../../datos/recetas/srep00196-s3.csv',
                    skip = 1, callback = ListCallback$new(limpiar))
recetas <-  filtrado %>% flatten

library(arules)
length(recetas)

[1] 2512

## No hacer mucho más chico que este soporte, pues tenemos relativamente
## pocas transacciones:
pars <- list(support = 0.05,  target = 'frequent itemsets',
             ext = TRUE)
ap_recetas <- apriori(recetas, parameter = pars)

Apriori

Parameter specification:

Algorithmic control:

Absolute minimum support count: 125 

set item appearances ...[0 item(s)] done [0.00s].
set transactions ...[242 item(s), 2512 transaction(s)] done [0.00s].
sorting and recoding items ... [41 item(s)] done [0.00s].
creating transaction tree ... done [0.00s].
checking subsets of size 1 2 3 4 5 6 done [0.00s].
sorting transactions ... done [0.00s].
writing ... [628 set(s)] done [0.00s].
creating S4 object  ... done [0.00s].

length(ap_recetas)

[1] 628

Vemos los items frecuentes

frecs <- ap_recetas %>% subset(size(.) == 1 ) %>% sort(by = 'support') %>%
 DATAFRAME
DT::datatable(frecs %>% mutate_if(is.numeric, function(x) round(x, 3)))

Registered S3 methods overwritten by 'htmltools':
  method               from         
  print.html           tools:rstudio
  print.shiny.tag      tools:rstudio
  print.shiny.tag.list tools:rstudio
Registered S3 method overwritten by 'htmlwidgets':
  method           from         
  print.htmlwidget tools:rstudio

Y ahora examinamos combinaciones frecuentes de distintos tamaños

ap_recetas %>% 
  subset(size(.) == 2) %>%
  subset(support > 0.20) %>%
  sort(by = 'support') %>%
  inspect

Incluso hay algunas combinaciones de 4 ingredientes que ocurren con frecuencia alta: estos ingredientes son bases de salsas, combinaciones de condimentos:

ap_recetas %>% 
  subset(size(.) == 4) %>%
  subset(support > 0.10) %>%
  sort(by = 'support') %>%
  inspect

pars <- list(support = 0.01, confidence = 0.10,
             target = 'rules',
             ext = TRUE)
reglas_recetas <- apriori(recetas, parameter = pars)

Apriori

Parameter specification:

Algorithmic control:

Absolute minimum support count: 25 

set item appearances ...[0 item(s)] done [0.00s].
set transactions ...[242 item(s), 2512 transaction(s)] done [0.00s].
sorting and recoding items ... [88 item(s)] done [0.00s].
creating transaction tree ... done [0.00s].
checking subsets of size 1 2 3 4 5 6 7 8 done [0.01s].
writing ... [50181 rule(s)] done [0.01s].
creating S4 object  ... done [0.01s].

agregar_hyperlift <- function(reglas, trans){
  quality(reglas) <- cbind(quality(reglas), 
    hyper_lift = interestMeasure(reglas, measure = "hyperLift", 
    transactions = trans))
  reglas
}
reglas_recetas <- agregar_hyperlift(reglas_recetas, recetas)

Análisis de pares comunes

library(arulesViz)

Loading required package: grid
Registered S3 method overwritten by 'seriation':
  method         from 
  reorder.hclust gclus
Registered S3 method overwritten by 'data.table':
  method           from
  print.data.table     

reglas_1 <- subset(reglas_recetas, hyper_lift > 1.1 & support > 0.1 & confidence > 0.40)
length(reglas_1)

[1] 213

reglas_tam_2 <- subset(reglas_1, size(reglas_1)==2)
#inspect(reglas_tam_2 %>% sort(by = 'hyper_lift')) 
plot(reglas_1 %>% subset(support > 0.2), engine = "plotly")

To reduce overplotting, jitter is added! Use jitter = 0 to prevent jitter.

library(tidygraph)

Attaching package: ‘tidygraph’

The following object is masked from ‘package:stats’:

    filter

library(ggraph)
frecs <- 
df_reglas <- reglas_tam_2 %>% DATAFRAME %>% rename(from=LHS, to=RHS) %>% data.frame
df_reglas$weight <- log(df_reglas$lift)
graph_1 <- as_tbl_graph(df_reglas) %>%
  mutate(centrality = centrality_degree(mode = "all")) 
set.seed(881)
ggraph(graph_1, layout = 'fr') +
  geom_edge_link(aes(alpha=lift), 
                 colour = 'red',
                 arrow = arrow(length = unit(4, 'mm'))) + 
  geom_node_point(aes(size = centrality, colour = centrality)) + 
  geom_node_text(aes(label = name), size=4,
                 colour = 'gray20', repel=TRUE) +
  theme_graph()

reglas_1 <- subset(reglas_recetas, hyper_lift > 1.5 & confidence > 0.1)
length(reglas_1)

[1] 11190

reglas_tam_2 <- subset(reglas_1, size(reglas_1)==2)
length(reglas_tam_2)

[1] 134

library(tidygraph)
library(ggraph)
df_reglas <- reglas_tam_2 %>% DATAFRAME %>% rename(from=LHS, to=RHS) %>% as_data_frame

`as_data_frame()` is deprecated, use `as_tibble()` (but mind the new semantics).
This warning is displayed once per session.

df_reglas$weight <- log(df_reglas$hyper_lift)
graph_1 <- as_tbl_graph(df_reglas) %>%
  mutate(centrality = centrality_degree(mode = "all")) 

ggraph(graph_1, layout = 'fr', start.temp=100) +
  geom_edge_link(aes(alpha=lift), 
                 colour = 'red',
                 arrow = arrow(length = unit(4, 'mm'))) + 
  geom_node_point(aes(size = centrality, colour = centrality)) + 
  geom_node_text(aes(label = name), size=4,
                 colour = 'gray20', repel=TRUE) +
  theme_graph()

Exportamos para examinar en Gephi:

write_csv(df_reglas %>% rename(source=from, target=to) %>%
            select(-count), 
          path='reglas.csv')

Nota

La combinación corn y starch puede deberse en parte a una separación incorrecta en el procesamiento de los datos (corn starch o maizena convertido en dos ingredientes, corn y starch):

df_reglas %>% filter(from == "{corn}", to == "{starch}")

La confianza es considerablemente alta, aunque tenemos pocos datos de esta combinación. Podemos examinar algunos ejemplos:

recetas %>% keep(~ "tomato" %in% .x & "corn" %in% .x) %>% head(10)

[[1]]
 [1] "tomato"        "vinegar"       "pork"          "celery_oil"    "leek"          "corn"          "black_pepper" 
 [8] "pepper"        "ginger"        "pea"           "garlic"        "soybean"       "soy_sauce"     "chicken_broth"
[15] "wine"         

[[2]]
 [1] "tomato"     "vinegar"    "pepper"     "celery_oil" "corn"       "cayenne"    "pork"       "garlic"    
 [9] "soybean"    "vegetable"  "coriander"  "rice"       "soy_sauce" 

[[3]]
[1] "tomato"     "vinegar"    "pork"       "celery_oil" "soy_sauce"  "ginger"     "garlic"     "sherry"    
[9] "corn"      

[[4]]
 [1] "pepper"        "celery_oil"    "starch"        "corn"          "ginger"        "garlic"        "soybean"      
 [8] "tomato"        "vinegar"       "beef"          "soy_sauce"     "cayenne"       "scallion"      "bell_pepper"  
[15] "vegetable_oil" "rice"          "wine"         

[[5]]
 [1] "tomato"     "vinegar"    "pork"       "celery_oil" "beef"       "soy_sauce"  "ginger"     "garlic"    
 [9] "corn"       "wine"      

[[6]]
 [1] "tomato"      "vinegar"     "pepper"      "lemon_juice" "celery_oil"  "sake"        "corn"        "pork"       
 [9] "ginger"      "honey"       "garlic"      "soybean"     "rice"        "soy_sauce"  

[[7]]
[1] "tomato"  "garlic"  "onion"   "bacon"   "corn"    "cayenne" "egg"    

[[8]]
 [1] "pork"              "green_bell_pepper" "celery_oil"        "starch"            "corn"             
 [6] "garlic"            "tomato"            "vinegar"           "onion"             "soy_sauce"        
[11] "cider"             "scallion"          "celery"            "pineapple"         "vegetable_oil"    
[16] "egg"              

[[9]]
 [1] "tomato"       "vinegar"      "pepper"       "celery_oil"   "roasted_pork" "soy_sauce"    "ginger"      
 [8] "honey"        "garlic"       "cinnamon"     "soybean"      "sherry"       "corn"         "oyster"      

[[10]]
 [1] "cane_molasses" "tomato"        "pork"          "celery_oil"    "vinegar"       "soy_sauce"     "pepper"       
 [8] "ginger"        "garlic"        "corn"